Fewer People, Different People, New Math: How the Biggest Companies in Tech Are Rebuilding Their Orgs From Scratch

In the span of a few weeks this spring, some of the largest and most watched companies in tech announced they were restructuring around the same idea: fewer people, completely different roles, and organizational models that would have been unrecognizable eighteen months ago.

Meta began cutting 8,000 employees while moving toward a 1:50 manager-to-engineer ratio, using internal AI agents to handle the administrative and project management work that once required thousands of middle managers. On a January earnings call, Mark Zuckerberg told investors he was starting to see projects that used to require big teams now accomplished by a single very talented person. He wasn't being abstract. He was describing how Meta was about to operate.

GitLab CEO Bill Staples announced what he called "Act 2," a restructuring for the "agentic era" that would flatten eight layers of management, double the number of small autonomous teams, and require daily AI usage from every remaining employee. Shopify's Tobi Lütke titled an internal memo "AI Usage is Now a Baseline Expectation" and made "prove you can't do it with AI before you ask for headcount" a standing policy for every team, himself and the executive team included. Klarna's Sebastian Siemiatkowski confirmed on the 20 VC podcast that the company had gone from 7,000 employees in 2022 to 3,000, with a target of 2,000 by 2030, almost entirely through AI automation and attrition.

And most recently, ClickUp CEO Zeb Evans cut 22 percent of his company and published a detailed thesis on X about what he's calling the "100x org." Agents outnumbering employees three to one. Product management and design merging into a single function. A new role called the "agent manager." Salary bands reaching $1 million in cash. **

Every one of these moves was framed as proactive. Every one was framed as structural rather than financial. And every one described an organization that looks nothing like the one it's replacing.

The question worth asking isn't whether these cuts are justified, but what the language around them reveals about how these companies are rethinking what an organization even is.

The redesign, not the reduction

A traditional layoff removes people from existing roles. You just have fewer humans filling the same boxes. What's happening now is different. The roles themselves are being declared obsolete, merged, or reinvented and the boxes are changing shape.

Zuckerberg is flattening management layers and replacing coordination overhead with internal AI agents. Staples is reorganizing GitLab's R&D into nearly sixty smaller teams with "end-to-end ownership" and naming agents as participants in reviews, approvals, and operational handoffs. Lütke is requiring every team to rethink its composition before adding a single human. Siemiatkowski is letting natural attrition hollow out an entire layer of the company while AI absorbs the work.

Evans went the most specific. He argued that product management and design are converging into a single "product builder" function. Designers with customer focus become more like PMs. PMs with UX intuition become more like designers. Agents handle the research and iteration cycles that used to justify entire teams. He went further with engineering, claiming that AI coding tools concentrate their productivity gains almost entirely in senior engineers, and that junior engineers using the same tools actually slow things down because the bottleneck shifts from writing code to reviewing code.

Whether any individual CEO's analysis is right matters less than the type of argument being made across the board. None of them are saying "we need to save money." They're saying "the unit of work has changed, which means the definition of a role has changed, which means the shape of the organization has changed." These are first-principles claims about what a company even is.

And the question of whether a role, as currently scoped and compensated, still describes a coherent unit of work is not limited to engineering. It applies to Brand Strategist and Creative Director just as easily. These CEOs are asking it across every function.

Three archetypes keep showing up

Across all of these companies, three role archetypes appear over and over. The names vary, but the shape of whats to come doesn't.

The Orchestrator. The senior person who directs systems and reviews their output. Evans calls this the "100x engineer." Zuckerberg described it as the single very talented person replacing a big team. They don't write more code or produce more content. They evaluate, steer, and make judgment calls at speed. What makes them valuable is taste, pattern recognition, the ability to know what "good" looks like and close the gap between what the machine produced and what the customer needs. GitLab's restructuring into small empowered teams is a bet on orchestrators. Shopify's headcount policy is the same logic applied from the demand side: before you hire a human executor, prove the orchestration layer can't handle it.

The System Owner. Evans calls this the "agent manager," someone who automates their own job, then becomes the owner and operator of the automated workflow. This is the most counterintuitive career move right now: the fastest way to become indispensable is to make yourself redundant at the task level. The person who understands the process deeply enough to encode it into a system, and then maintains, improves, and troubleshoots that system, holds institutional knowledge that the company cannot replace. They took the work and turned it into infrastructure. Meta's move toward AI agents handling administrative and project management tasks is creating this role at enormous scale.

The Human Bottleneck (by design). Evans explicitly protected customer-facing roles from the automation push. One-on-one meeting time with customers, he said, is a bottleneck that should not be replaced "even when agents are high enough quality to do video meetings." The systems around those meetings (scheduling, preparation, follow-up, analysis) get fully automated. The meeting itself stays human. Siemiatkowski landed on the same conclusion the hard way. Klarna replaced 700 customer service agents with AI, watched customer satisfaction fall, and is now rebuilding its human support capacity. "I just think it's so critical that you are clear to your customer that there will be always a human if you want," he said.

The implications of that last category go further than the others. When most interactions are automated, the remaining human touchpoints become the luxury good, the differentiator, the thing competitors can't replicate by spinning up another agent. These companies aren't eliminating humans from the equation. But the distinction is that they're concentrating human presence at the points where it matters most and automating the rest.

The salary band question

ClickUp's $1 million salary bands are the most provocative compensation signal in this wave, but the logic behind them shows up everywhere.

Meta has offered pay packages worth hundreds of millions of dollars over four years to court top AI researchers for its new superintelligence team. The companies making the deepest cuts are simultaneously making the most aggressive compensation offers to the people they want to keep or recruit. The math isn't complicated. If one person with AI leverage produces the output that previously required ten people, you can pay them five times their old salary and still cut your labor costs in half. The surplus has to go somewhere: to shareholders, to infrastructure investment, or back to the remaining employees.

This flips how most companies think about salary bands. Traditional compensation is anchored to role, level, and market rate, all proxies for expected contribution. What Evans and Zuckerberg and Lütke are all describing, in different vocabularies, is compensation anchored to actual leverage. How much value does this specific person, with these tools, actually produce? The role and the level become irrelevant and the output is everything. 

That logic doesn't stop at engineering. What does a "100x marketer" look like? A "100x creative director?" The concept might land differently in a domain where the value lives in judgment, taste, and relationships that resist easy measurement. Or it might land exactly the same way, just with different tools. Either way, when companies in your talent market start paying seven figures for leverage-based output, every other company's salary bands start looking like relics.

The restructuring as narrative

One of the most telling details in all of this is where and how these announcements are being made.

Evans posted his restructuring as a public thesis on X, complete with a framework, a philosophy, and new organizational vocabulary. Lütke posted his internal memo to X after it started leaking, then let it become a recruiting signal. Siemiatkowski has been openly narrating Klarna's headcount decline on podcasts and in press interviews for over a year. GitLab published its restructuring plan as a blog post. Zuckerberg described the new Meta on an earnings call.

These companies are treating their organizational redesigns as public communications. The restructuring announcement is functioning as a signal on three levels at once: to customers ("we're faster and more capable"), to investors ("we're efficient and forward-looking"), and to talent ("the best people want to work here, and we'll pay them accordingly").

When ClickUp says "$1 million salary bands" and Meta offers nine-figure packages to AI researchers, those are also recruiting magnets aimed at the exact people these companies need. The companies doing this well will attract the orchestrators and system owners. The companies not doing it at all will watch that talent pool flow elsewhere.

The tension at the center

Every one of these CEOs uses remarkably similar language. "This isn't about cost cutting." "Most savings will be reinvested." "This is proactive, not reactive." Evans, Staples, Lütke, Siemiatkowski, Zuckerberg. The consistency is notable enough to raise the question: is it true?

Some observers have started calling it "AI washing," using the vocabulary of transformation to provide narrative cover for what would otherwise be a margin play. The counterargument is that the restructuring details (new role definitions, merged functions, agent-to-human ratios, flattened hierarchies, 1:50 manager ratios) are too specific and too structural to be theater. You don't invent a new job title called "agent manager" and publish a thesis about why junior engineers slow down senior engineers just to dress up a cost cut. You don't move to a 1:50 manager-to-engineer ratio unless you actually believe the coordination layer can be automated.

The financial incentive to cut headcount is real and needs no AI justification. The structural change in how work gets done is also real and would demand organizational redesign even if margins were fine. Both things are happening at once. The companies that treat this as only a cost exercise, cutting people without redesigning the work itself, will end up with a smaller team doing the same jobs the same way, just under more pressure. That's not a 100x org. That's a skeleton crew. And Klarna already showed what happens when you automate too aggressively without redesigning around the gaps.

The companies that actually redesign, that take seriously the question of what roles should exist, what those roles should be paid, and what the relationship between humans and machines should look like, will end up somewhere different. Whether it's better remains to be seen. But different compounds.

The 100x org as blueprint

Zuckerberg is betting $135 billion on this shift. Siemiatkowski is narrating the shrink in real time. Lütke has made it company policy. Staples is rebuilding GitLab from the ground up around it.

But Evans may have produced the most detailed public blueprint of what one of these redesigned organizations actually looks like in practice. His "100x org" framework lays out the whole thing: which roles survive, which ones merge, which new ones get created, how compensation changes, and what the ratio of humans to agents looks like on the other side. He's specific about what product builders do versus system owners versus customer-facing front-liners. He's specific about why the productivity gains concentrate in senior people and why more code from junior engineers creates review bottlenecks rather than speed. He's specific about where human presence stays protected and why.

You can disagree with any of it. You can think the "100x" framing is aspirational marketing for what is, at the end of the day, a 22 percent cut at a $4 billion company eyeing an IPO. But the specificity matters. Most restructuring announcements gesture vaguely at "AI transformation" and "operational efficiency." Evans published an org design thesis, role by role, with a compensation philosophy attached. That's a different kind of document. It's closer to an open-source operating manual for what a company looks like when you stop bolting AI onto the existing structure and start rebuilding the structure around what AI makes possible.

"Nearly every company will make changes like these," Evans wrote. "The ones that do it proactively will define what comes next."

Some of these bets will work. Some won't. But the direction is set. Roles are being merged. Salary bands are being blown open. Management layers are being removed. Entire functions are being redefined around what humans do versus what machines do. The era of treating organizational design as a static backdrop to strategy, rather than the strategy itself, is over. Every company is deciding what their org chart looks like right now, whether they realize it or not.

** The author of this article and SoB's editor in chief, Melissa Rosenthal, was previously the Chief Creative Officer at ClickUp.